HAL Id: inserm-01187746
https://www.hal.inserm.fr/inserm-01187746
Submitted on 27 Aug 2015
HAL is a multi-disciplinary open access archive for the deposit and dissemination of sci- entific research documents, whether they are pub- lished or not. The documents may come from teaching and research institutions in France or abroad, or from public or private research centers.
L’archive ouverte pluridisciplinaire HAL, est destinée au dépôt et à la diffusion de documents scientifiques de niveau recherche, publiés ou non, émanant des établissements d’enseignement et de recherche français ou étrangers, des laboratoires publics ou privés.
Mapping Integrity of white matter microstructure in alcoholics with and without Korsakoff’s syndrome
Shailendra Segobin, Ludivine Ritz, Coralie Lannuzel, Céline Boudehent, François Vabret, Francis Eustache, Hélène Beaunieux, Anne Lise Pitel
To cite this version:
Shailendra Segobin, Ludivine Ritz, Coralie Lannuzel, Céline Boudehent, François Vabret, et al.. In- tegrity of white matter microstructure in alcoholics with and without Korsakoff’s syndrome Human Brain Mapping Integrity of white matter microstructure in alcoholics with and without Korsakoff’s syn- drome. Human Brain Mapping, Wiley, 2015, pp.2795-2808. �10.1002/hbm.22808�. �inserm-01187746�
For Peer Review
Integrity of white matter microstructure in alcoholics with and without Korsakoff’s syndrome
Journal: Human Brain Mapping Manuscript ID: HBM-14-1178.R2 Wiley - Manuscript type: Research Article Date Submitted by the Author: n/a
Complete List of Authors: Segobin, Shailendra; Inserm-EPHE-CHU-Université de Caen/Basse- Normandie, Unité U1077,
Ritz, Ludivine; Cyceron,
Lannuzel, Coralie; Inserm-EPHE-CHU-Université de Caen/Basse- Normandie, Unité U1077,
Boudehent, Céline; Inserm-EPHE-CHU-Université de Caen/Basse- Normandie, Unité U1077,
Vabret, François; Inserm-EPHE-CHU-Université de Caen/Basse-Normandie, Unité U1077,
Eustache, Francis; Inserm-EPHE-CHU-Université de Caen/Basse- Normandie, Unité U1077,
Beaunieux, Hélène; Inserm-EPHE-CHU-Université de Caen/Basse- Normandie, Unité U1077,
Pitel, Anne-Lise; Inserm, EPHE, Université de Caen/Basse-Normandie, Unité U1077, GIP Cyceron, CHU de Caen
Keywords: Alcoholism, Korsakoff's syndrome, Tract-Based Spatial Statistics, Classification, Frontocerebellar circuit, Papez Circuit
For Peer Review
Integrity of white matter microstructure in alcoholics with and without
1
Korsakoff’s syndrome
2
Shailendra Segobin1,2,3,4 , Ludivine Ritz1,2,3,4, Coralie Lannuzel1,2,3,4 , Céline Boudehent1,2,3,4,5, François 3
Vabret1,2,3,4,5, Francis Eustache1,2,3,4, Hélène Beaunieux1,2,3,4 and Anne Lise Pitel1,2,3,4 4
5
1 INSERM, U1077, Caen, France 6
2 Université de Caen Basse-Normandie, UMR-S1077, Caen, France 7
3 Ecole Pratique des Hautes Etudes, UMR-S1077, Caen, France 8
4 Centre Hospitalier Universitaire, Caen, France 9
5 Centre Hospitalier Universitaire, service d’addictologie, Caen, France 10
11 12
Corresponding author: AL Pitel 13
GIP Cyceron, Bd Becquerel, 14
Caen Cedex 14074, France 15
Tel.: +33 (0)2 31 47 02 47;
16
Fax: +33(0)2 31 47 02 75;
17
E-mail: pitel@cyceron.fr 18
19 20 21 2
3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59
For Peer Review
Abstract 22
Alcohol dependence results in two different clinical forms: “uncomplicated” alcoholism (UA) 23
and Korsakoff’s syndrome (KS). Certain brain networks are especially affected in UA and 24
KS: the frontocerebellar circuit (FCC) and the Papez circuit (PC). Our aims were 1) to 25
describe the profile of white matter (WM) microstructure in FCC and PC in the two clinical 26
forms, 2) to identify those UA patients at risk of developing KS using their WM 27
microstructural integrity as a biomarker.
28
Tract-based spatial statistics and non-parametric voxel-based permutation tests were used to 29
compare DTI data in 7 KS, 20 UA and 14 healthy controls. The two patient groups were also 30
pooled together and compared to controls. k-means classifications were then performed on 31
mean FA values of significant clusters across all subjects for 2 fiber tracts from the FCC (the 32
middle cerebellar peduncle and superior cerebellar peduncle) and 2 tracts from the PC (fornix 33
and cingulum).
34
We found graded effects of WM microstructural abnormalities in the PC of UA and KS. UA 35
patients classified at risk of developing KS using fiber tracts of the PC also had the lowest 36
scores of episodic memory. That finding suggests that WM microstructure could be used as a 37
biomarker for early detection of UA patients at risk of developing KS.
38
Keywords: alcoholism, Korsakoff’s syndrome, Fronto-cerebellar circuit, Papez’s circuit, 39
TBSS, classification 40
41
42
43 2
3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59
For Peer Review
Introduction
44
The effects of chronic and excessive alcohol consumption on the human brain and 45
cognition result, among others, in two different clinical forms of alcohol-dependence, which 46
differ mainly by the extent of brain damage (Pitel et al. 2012). The more severe clinical form 47
is the Korsakoff’s syndrome (KS, Korsakoff, 1889), which is defined by permanent and 48
debilitating neurological complications that arise from a combination of heavy alcohol 49
consumption and thiamine deficiency. Patients with KS suffer from retrograde (Kopelman.
50
1989; for review Oscar-Berman. 2012) and anterograde (Fama et al. 2012 for review) 51
amnesia, as well as ataxia (Sullivan et al. 2000), visuospatial deficits (Jacobson et al. 1990) 52
and executive dysfunctions (Oscar-Berman. 2012 for review). Postmortem (Harper and Kril.
53
1990; Harper, 2009; Mayes et al. 1988; Victor et al. 1971) and neuroimaging studies 54
(Colchester et al. 2001; Krabbendam et al. 2000; Pitel et al. 2009; Sullivan and Marsh, 2003;
55
Sullivan et al. 1999) revealed structural brain abnormalities in KS patients, especially in the 56
thalamus, cingulate cortex, cerebellum, mammillary bodies and white matter tracts of the 57
superior cerebellar vermis (Harper et al. 2003). Studies using magnetic resonance imaging 58
(MRI) have also shown shrinkage of the frontal and parietal cortices (Christie et al. 1988).
59
The impact of the pathology on the hippocampus is still under debate with some studies 60
reporting the region as preserved (Colchester et al. 2001; Squire et al. 1990), while others 61
found it damaged (Sullivan and Marsh, 2003; Visser et al. 1999) 62
The other clinical form of alcohol-dependence refers to patients often considered as 63
“Uncomplicated Alcoholics” (UA, Pitel et al. 2009, 2012; for review Oscar Berman et al, 64
2014; Zahr, 2014). Those patients, equally coined as “detoxified alcoholics” (Chanraud et al.
65
2007 for example), or “non-Korsakoff alcoholics” (Parsons, 1998 for example) are those 66
without ostensible and severe neurological complications or liver dysfunctions (Alexander- 67
2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59
For Peer Review
Kauffman et al., 2006, Harper, 2007; Harper and Matsumoto, 2005; Matsumoto 2009; for 68
review Oscar-Berman et al, 2014; Zahr, 2014). This clinical form is characterized by its 69
heterogeneity but is known to result in mild-to-moderate cognitive deficits (Parsons and 70
Nixon. 1998; Sullivan et al. 2000) and brain damage (Chanraud et al. 2007; Rosenbloom et al.
71
2003). The neuropsychological profile includes impairment of working memory and 72
executive functions such as planning, organisation, categorisation, flexibility, inhibition, and 73
deduction of rules (Ambrose et al. 2001; Ihara et al. 2000; Noël et al. 2001; Pitel et al. 2007).
74
Episodic memory is also affected in UA with both encoding and retrieval processes being 75
impaired (Noël et al. 2012; Pitel et al. 2007). Neuroimaging investigations revealed gray 76
matter volume losses in the frontal, parietal and medial temporal lobes, the cerebellar cortex, 77
cerebellar vermis, as well as subcortical structures including the thalamus and the caudate 78
nucleus (Chanraud et al. 2007; Pfefferbaum et al. 1992; Shear et al. 1996; Sullivan. 2003;
79
Sullivan et al. 2003). Neuropathological studies reported a disruption of the cytoskeleton and 80
white matter shrinkage especially in the corpus callosum, superior frontal cortex, anterior 81
superior cerebellar vermis and the limbic system (Chanraud et al. 2009; Harris et al. 2008;
82
Pfefferbaum et al. 2006, 2009). While post-mortem examination of the human brain did not 83
indicate demyelination because of the inherent rapid disintegration of cellular membranes 84
following death, animal studies have shown thinning of myelin sheaths (Phillips et al. 1991).
85
Differences in microstructural integrity have also been found in-vivo in diffusion tensor 86
imaging (DTI) studies of alcoholic patients having lower fractional anisotropy (FA) in the 87
genu of the corpus callosum in men and the centrum semiovale in women (Pfefferbaum and 88
Sullivan, 2005). When WM fibre tracts were defined a priori, lower FA values were revealed 89
in the mesencephalic and pontine region, superior longitudinal fasciculus, external capsule, 90
fornix and cingulum (Chanraud et al. 2009; Harris et al. 2008; Pfefferbaum et al. 2006, 2009).
91 2
3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59
For Peer Review
In both UA and KS, two brain networks and associated cognitive functions are 92
predominantly affected: the frontocerebellar circuit (Chanraud et al. 2010) and the Papez 93
circuit (Aggleton. 2012; Parsons. 1998). The frontocerebellar circuit (FCC), identified in non- 94
human primates using viral transneuronal tracing technology (Kelly and Strick, 2003), 95
consists of two distinct, parallel closed-loops within the cortico-thalamo-cerebellar circuitry.
96
The first one underlies executive functions and includes Brodmann areas 9 and 46 of the 97
dorsolateral prefrontal cortex, which receives input from the cerebellar crus I and II through 98
the thalamus, and projects back to the cerebellum through the pons. The second one 99
contributes to motor functions and encompasses the motor cortex, which receives input from 100
lobules IV-VI of the cerebellar vermis through the thalamus and feeds back to the cerebellum 101
via the pons (Chanraud et al. 2010). Diffusion tensor imaging technique and tractography 102
analyses (Mori et al. 2010) have shown that the superior cerebellar peduncle connects the 103
cerebellum to the thalamus, which is then connected to the dorsolateral prefrontal cortex 104
through the anterior limb of the internal capsule and the anterior corona radiata. The feedback 105
loop goes from the dorsolateral prefrontal cortex back to the pons via corticopontine tracts 106
and back to the cerebellum through the middle cerebellar peduncle. The Papez circuit (PC, 107
Aggleton and Brown, 1999; Papez, 1937), involved in episodic memory, includes the 108
hippocampal formation, which connects to the mammillary bodies via the fornix. The 109
thalamus then receives information from the mammillary bodies via the mammilo-thalamic 110
tract. The anterior limb of the internal capsule connects the thalamus to the cingulate gyrus, 111
which is in turn connected back to the hippocampus through the cingulum bundle.
112
The PC and the FCC are affected in both clinical forms, but not to the same extent, 113
hence the difference of clinical severity between UA and KS patients. In fact, the comparison 114
between these two clinical forms of alcohol-dependence revealed patterns of differences and 115
similarities in the profiles of cognitive impairments and brain structural deficits.
116 2
3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59
For Peer Review
Neuropsychological and neuroimaging studies have shown that impairment of the FCC is 117
generally comparable in both UA and KS, whereas graded effects are observed for the PC 118
(Pitel et al. 2008, 2012). More precisely, it has been shown that while deficits in working 119
memory and executive functions did not differ significantly between UA and KS patients, 120
deficits in episodic memory are more severe in KS patients compared to UA (Butters and 121
Brandt. 1985; Fama et al. 2012; Pitel et al. 2008). Volumetric analyses of gray matter (Harper, 122
2009; Harper et al. 2003; Pitel et al. 2009, 2012; Sullivan and Pfefferbaum, 2009) have also 123
indicated that nodes belonging to the PC including the medial thalami and mammillary bodies 124
were more severely affected in KS compared to UA patients. Similar degrees of shrinkage 125
have been observed in both patient groups in some of the nodes of the FCC including the 126
frontal cortex but not in the pons and cerebellum, for which graded effects were observed 127
(Sullivan and Pfefferbaum, 2008). The few studies that have compared UA and KS patients 128
regarding white matter volumes indicated that the cerebellar white matter (Kril et al. 1997), 129
corpus callosum and thalamic radiations were more severely damaged in KS than in UA 130
(Harper et al. 2003; Pitel et al. 2012).
131
The comparison of neuropsychological functioning between UA and KS patients gave 132
rise to the hypothesis that the effects of chronic alcohol consumption lie along a continuum 133
from mild-to-moderate impairments in UA to severe ones in patients with KS (Butters and 134
Brandt, 1985; Parsons, 1998; Ryback, 1971). The existence of a continuum between these two 135
clinical forms reflects the heterogeneity within the UA group with some patients having 136
preserved results similar to those of healthy controls, while others have severe deficits close to 137
those of KS patients (Pitel et al. 2008). The latter ones are at risk to develop severe alcohol- 138
related neurological complications but they often go undiagnosed (Pitel et al. 2011) and 139
therefore do not receive appropriate treatment. Early identification of these patients would 140
help clinicians in optimising treatment outcome. Clinical, neuropsychological and 141
2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59
For Peer Review
macrostructural brain biomarkers of risks for KS have been proposed. Previous studies have 142
suggested that a subgroup of UA patients with episodic memory impairments (Pitel et al.
143
2008) and thalamic shrinkage (Pitel et al. 2012) close to those of KS patients can be 144
identified. This subgroup of patients may even be clinically defined by the presence of signs 145
of Wernicke’s encephalopathy (Pitel et al. 2011). How white matter microstructure can also 146
be used as a biomarker of KS has never been explored. The white matter microstructure and 147
structural connectivity has never been investigated in KS and therefore the integrity of the 148
fibres bundles has never been quantitatively compared between UA and KS. Comparisons of 149
white matter have so far been limited to neuropathological investigations (Harper, 2009) and 150
one voxel-based morphometry (VBM) study of white matter volumes (Pitel et al. 2012).
151
Those studies gave first insights regarding white matter volumes but did not provide a clear 152
picture of the differences in the microstructure of white matter bundles. The latter is better 153
represented by observing the differences in FA values of white matter fibre tracts, obtained 154
through DTI studies, under the assumption that FA is a structural biomarker that depicts white 155
matter disruption involving myelin, cytoskeleton and the axons’ microtubule system 156
(Pfefferbaum et al. 2006). We should however bear in mind that in absence of sound 157
methodological procedures, measurements of FA values could turn out to be artefactual 158
instead of reflecting impairments due to factors inherent to alcohol-dependence. One 159
example is the ‘correspondence problem’ (Smith et al. 2006) following poor spatial 160
normalization. A preserved white matter tract would be observed as impaired if what 161
are being effectively measured are FA values in crossing fibres which are inherently 162
lower.
163
The first objective of the present study was therefore to describe the white matter 164
microstructure in UA and KS compared to healthy controls (HC) using a voxel-wise 165
approach. We hypothesize graded effects of compromised white matter integrity in the 166
2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59
For Peer Review
bundles of the PC (KS<UA<HC) but not in those of the FCC ((KS=UA)<HC). Since the UA 167
group is classically heterogeneous, the second objective was to identify UA patients at risk of 168
developing KS through the analysis of white matter integrity across the tracts belonging to the 169
FCC and PC.
170
Materials and Methods
171
Participants
172
Twenty-seven patients (22 men, 5 women) with alcohol-dependence (DSM IV criteria, 173
American Psychiatric Association. 1994) and 14 healthy subjects (9 men, 5 women) were 174
included in the study. To be included, all participants had to be between 18 and 70 years old, 175
and to have French as their native language. No participant had a comorbid psychiatric 176
disorder (no other axis 1 of the DSM IV as evaluated by MINI 500, American Psychiatric 177
Association 2004), was under psychotropic medication, had a history of serious chronic 178
pathology (diabetes, hepatitis, HIV, endocrinal disorder, as revealed by participants’ blood 179
tests), neurological problems (traumatic head injury causing loss of consciousness for >30 180
minutes, epilepsy, stroke, etc.) that might have affected cognitive function. No participant 181
fulfilled the DSM-IV criteria for abuse of another substance over the last 3 months, nor filled 182
the DSM-IV criteria for dependence of another substance (except tobacco). They had not 183
taken any other psychoactive substance for more than 5 times over the last month (except 184
alcohol for the patients) and had not participated in any neuropsychological study or had any 185
neuropsychological evaluation during the previous year. All participants gave their informed 186
consent to the study, which was approved by the local ethics committee. Their demographical 187
details are summarised in Table 1. All patients were recruited for this study while being 188
inpatients at Caen University Hospital. The study was carried out in line with the Declaration 189
of Helsinki (1964).
190 2
3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59
For Peer Review
Of those 27 patients, 7 (6 men, 1 woman) filled the DSM IV criteria of persisting 191
amnestic disorder (American Psychiatric Association. 1994) and were therefore diagnosed as 192
KS patients. All KS patients had a history of heavy drinking (longer than 20 years), a Mini- 193
Mental State (MMS, Folstein, 1975) score of at least 20, were abstinent for at least 7 days and 194
were diagnosed with severely impaired episodic memory as revealed by a neuropsychological 195
examination. The consequences of their memory impairments were such that none of the KS 196
were able to go back to their previous jobs and all of them lived in sheltered accommodation 197
or were inpatients waiting for a place in an institution. It was difficult to obtain accurate 198
information about their alcohol intake due to their amnesia. The background information for 199
the KS came mainly from family members and medical records. For each KS patient, the 200
selection was made according to a codified procedure in a French officially registered centre 201
for addiction. The case of each patient was examined by a multidisciplinary team made up of 202
specialists in cognitive neuropsychology and behavioural neurology. Clinical and 203
neuroimaging investigations ruled out other possible causes of memory impairments 204
(particularly focal brain damage).
205
The 20 alcoholic patients without KS were considered as UA patients. They were 206
recruited by clinicians while being inpatients for alcohol-dependence at Caen University 207
Hospital. Although patients were early in abstinence (2.4±3.1 days of sobriety prior to 208
inclusion), none of them presented physical symptoms of alcohol withdrawal as assessed by 209
the Cushman’s scale (Cushman et al. 1985) at inclusion. They were interviewed with the 210
Alcohol Use Disorders Identification Test (AUDIT; (Gache et al. 2005)) and a modified 211
version of the semi-structured lifetime drinking history (Pfefferbaum et al. 1988). Measures 212
included the duration of alcohol use (in years), alcohol misuse (in years), alcohol dependence 213
(in years), number of withdrawal and daily alcohol consumption prior to treatment (in units, a 214
standard drink corresponding to a beverage containing 10 g of pure alcohol).
215 2
3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59
For Peer Review
The control group (HC) was recruited locally mainly by word of mouth and to match 216
the demographics of the UA patients. Inclusion criteria were: a minimum MMS score of 26 or 217
a minimum MATTIS (Mattis, 1976) score of 129, and a maximum Beck Depression Index 218
(Beck et al. 1961) of 29. The maximum score at the Alcohol Use Disorders Test (AUDIT) 219
was 6 for women and 7 for men.
220
UA and HC were age- and education-matched (p=0.723 and p=0.76 respectively). KS 221
differed from both HC and UA in age, education (years of schooling) and MMSE scores. Age, 222
education, depression (Beck Depression Inventory, and anxiety scores (State-Trait Anxiety 223
Inventory (STAI) for adults with two forms Y-A for “state anxiety” and Y-B for “trait 224
anxiety”) (Spielberger et al. 1983) as well as nicotine dependence level (Fagerstrom Test, 225
(Heatherton et al. 1991)) are reported in Table 1.
226
All participants underwent a neuropsychological examination assessing intellectual 227
abilities (Information and Matrix Reasoning subtests of the WAIS III (Wechsler, 2001a)), 228
global cognitive function (MMSE; Folstein et al. 1975) and episodic memory (the French 229
version of the Free and Cued Selective Reminding Test FSCRT (Grober and Buschke. 1987;
230
Van der Linden, 2004). Neuropsychological performances are reported in Table 2.
231
DTI data acquisition
232
All participants underwent a DTI sequence on the Philips Achieva 3T MRI scanner 233
(Netherlands). 70 slices (slice thickness of 2mm, no gap) were acquired axially using a 234
diffusion weighted imaging spin echo (DWI-SE) sequence (32 directions at b=1000 s/mm2, 235
repetition time = 10000 ms; echo time = 82 ms; flip angle = 90°, field of view = 224x224 236
mm2; matrix= 112 x 112 and in-plane resolution of 2x2 mm2; one no-diffusion weighted 237
image at b=0 s/mm2 was also acquired).
238 2
3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59
For Peer Review
DTI data processing
239
The diffusion-weighted images (DWI) for all subjects were first pre-processed to create FA 240
images using the FSL Diffusion Toolbox (FDT) (http://fsl.fmrib.ox.ac.uk/fsl/fslwiki/FDT) 241
that is part of FSL 5.0 toolbox for medical image analysis (Smith et al. 2004). The FA images 242
were further processed using Tract Based Spatial Statistics (TBSS) for subsequent voxelwise 243
statistical analysis (Smith et al. 2006). TBSS presents an improvement on classical voxelwise 244
approaches like voxel based morphometry (VBM). More specifically, it first addresses the 245
“correspondence problem” faced by standard registration algorithms where it is difficult to 246
gauge whether the observed differences are indeed due to differences in tissue 247
volumes/density or are artefactual modifications that result from local misalignment. The 248
issue becomes more pertinent in the case of white matter tracts where a higher level of 249
precision is required to ensure that the FA values contained in the voxels come from exactly 250
the same part of WM tract across all subjects. TBSS addresses the problem by tailoring the 251
non-linear registration algorithm to the requirements of the DTI data, followed by projection 252
onto a tract representation that is an alignment invariant (referred to as the mean FA skeleton).
253
Such an approach also removes the need for applying a spatial smoothing for which the 254
choice of the smoothing kernel is deemed to be done in an arbitrary manner, and results 255
known to be highly dependent on the kernel size (Jones et al. 2005). Smoothing increases 256
partial volume effects between tissues such that it is difficult to differentiate between WM 257
differences that are due to the biological mechanism under investigation or an artefactual 258
measurement due to a mixture of tissues.
259
For each subject, the 32 DWI images were first corrected for distortions due to Eddy currents 260
and aligned to the b=0 s/mm2 image using rigid-body registration for motion correction 261
(Jenkinson et al. 2002). FA images were then created by fitting a tensor model to the diffusion 262
images and were further processed using TBSS. Briefly, all subjects' FA data were aligned 263
2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59
For Peer Review
into MNI space using the nonlinear registration tool (FNIRT), which uses a b-spline 264
representation of the registration warp field (Rueckert et al. 1999) resulting in FA maps of 265
matrix size of 182x218x182 and voxel size of 1x1x1 mm3. Next, the mean FA image was 266
calculated and thinned to create a mean FA skeleton, which represents the centres of all tracts 267
common to the group. Each subject's aligned FA image was threshold at 0.3 to exclude low 268
FA values that could be contaminated with partial volume effects from other non-white- 269
matter tissues and to minimise inter-subject variability. The resulting image is then projected 270
onto the mean skeleton by filling every voxel of the latter with the maximum FA value that 271
lies perpendicular to the skeleton structure. Voxel-based statistics are performed on these 272
‘skeletonised’ images.
273
Statistical analyses
274
A. Comparison of white matter integrity in HC, UA patients and KS patients in the 275
whole-brain (voxel-based analysis) 276
Non-parametric permutation tests (Nichols and Holmes. 2002) were performed between HC 277
and UA; HC and KS; and UA and KS groups. Age was included as a covariate to account for 278
between-group differences (KS being older than the two other groups). For each between- 279
group comparison, 5000 permutations were done, and the data corrected for multiple 280
comparisons (FWE, p<0.05) using threshold-free cluster enhancement (TFCE; (Smith and 281
Nichols. 2009)) for cluster-wise correction. This statistical toolbox is implemented within 282
FSL 5.0. (http://fsl.fmrib.ox.ac.uk/fsl/fslwiki/Randomise).
283
B. Identification of UA patients at risk of developing KS 284
1. Because of the heterogeneity within the UA group, UA and KS patients were pooled 285
together to form a single group of alcohol-dependant patients (UASK). The UASK group 286
2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59
For Peer Review
was then compared to the HC group using non-parametric permutation tests (5000 287
permutations, FWE p<0.05, TFCE).
288
2. The John Hopkins University International Consortium for Brain Mapping (JHU-ICBM) 289
DTI-81 WM atlas (Mori et al. 2010), implemented as an atlas tool in FSL 5.0 290
(http://fsl.fmrib.ox.ac.uk/fsl/fslwiki/Atlases), was then used to extract and binarise clusters 291
highlighted in the previous between-group analysis (UASK vs HC). Only the WM fibre 292
tracts that belong to the FCC and PC were used to produce “fibre-cluster masks”. These 293
“fibre-cluster masks” were then employed to extract the mean FA value within each tract 294
belonging to the FCC and PC for each subject. Focus was laid on a couple of fibre tracts 295
only for each circuit. In that respect, the cingulum and the fornix were chosen for the PC 296
and the middle cerebellar peduncle and superior cerebellar peduncle for the FCC. These 297
are the tracts that are believed to offer less variability in terms of specificity to the PC and 298
FCC respectively, having a lower involvement in fibres connecting other parts of the brain 299
or involved in other circuitry.
300
3. For each fibre tract, a k-means clustering classification was then performed on the mean 301
FA values, with the algorithm constrained to separate the 41 participants into 2 groups.
302
The aim was to find which white matter fibre tract enables the identification of some UA 303
patients classified into the same group as KS patients. A reliable classification would 304
include all HC into one group, and all KS patients into the other group with the 305
heterogeneous UA group fitting into either of these. Those UA who will be sorted within 306
the same group as the KS patients will be deemed as UA_HIGH patients (for high risk of 307
developing KS) while those who will be sorted within the same group as the HC will be 308
labelled UA_LOW patients (for low risk of developing KS). It is hypothesised that a more 309
robust classification will be obtained when using FA values in the fibres of the PC than in 310
2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59
For Peer Review
those of the FCC. Comparisons of episodic memory performance among the subsequent 311
subgroups should help in asserting the robustness of the classification step.
312
4. Between-group comparisons of episodic memory performance:
313
Episodic memory scores were compared between the groups of HC, UA_LOW, 314
UA_HIGH and KS using non-parametric Mann Whitney U tests. The hypothesis is that 315
the scores on the episodic memory test in UA_LOW patients will differ significantly from 316
the UA_HIGH and KS patients but not from HC. UA_HIGH patients are expected to 317
differ significantly from HC and UA_LOW and KS patients.
318
Results
319
A. Comparison of white matter integrity in HC, UA patients and KS patients
320
(i) UA versus HC 321
Non-parametric permutation tests (FWE, p<0.05) between UA and HC showed lower FA 322
values in fibre tracts spread across the whole brain including the corpus callosum (Tmax = 323
6.01; k = 6258; η2= 0.17), the anterior limb of the internal capsule (Tmax = 5.86; k=742;
324
η2=0.10), the anterior corona radiata (Tmax = 5.52; k=2215; η2=0.16), the fornix (Tmax = 5.91;
325
k=560; η2= 0.15); the cingulum (Tmax = 4.95; k=821; η2= 0.14); the middle cerebellar 326
peduncle (Tmax = 5.17; k=1320; η2=0.11) and the superior cerebellar peduncle Tmax = 4.92;
327
k=252; η2=0.15) as shown in Fig. 1.
328
Note: Values for effect size (η2) have been calculated from T-scores resulting from the non- 329
parametric permutation tests.
330
(ii) KS versus HC 331
The profile of WM structural impairment in KS versus HC was similar to that between UA 332
and HC. FA values were lower in KS than in HC in the same tracts as reported above but with 333
higher T and k values: corpus callosum (Tmax = 10.52; k=7279; η2=0.26) , anterior corona 334
2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59
For Peer Review
radiata (Tmax = 7.45; k=2936; η2=0.31), anterior limb of the internal capsule (Tmax = 7.00;
335
k=1179; η2=0.23); cingulum (Tmax = 6.93; k=869; η2=0.14); the fornix (Tmax = 6.60; k=693;
336
η2=0.31); the middle cerebellar peduncle (Tmax = 6.30; k=1153; η2=0.27) and the superior 337
cerebellar peduncle Tmax = 7.27; k=270; η2=0.35).
338
(iii) UA versus KS 339
Group differences between UA and KS were found mainly in the corpus callosum (Tmax = 340
4.93; k=4453; η2=0.139) and anterior corona radiata (Tmax = 4.9; k=932; η2=0.13). Other 341
tracts with significant differences were also observed but they either had low T-values or their 342
cluster sizes were small. For example, anterior limb of the internal capsule (Tmax = 3.59;
343
k=163; η2=0.11); cingulum (Tmax = 3.61; k=103; η2=0.14) and the fornix (Tmax = 3.81; k=107;
344
η2=0.16). There were no significant differences in the middle and superior cerebellar 345
peduncles.
346
[FIGURE 1]
347
There was no gender effect in the UA and KS groups across all white matter fibre tracts. In 348
the control group, differences between men and women were significant for the cingulum 349
(Mann-Whitney U test, p=0.014). There was no correlation between FA measures and BDI 350
scores or years of education across all tracts for all 3 groups. There was also no correlation 351
between FA measures and the number of detoxifications in the UA group. The same analysis 352
could not be performed for the KS group since it was not possible to obtain complete and 353
accurate information regarding alcohol history for these patients.
354
355
356
357 2
3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59
For Peer Review
B. Identification of UA patients at risk of developing KS
358
(i) Comparison of white matter integrity between HC and UASK 359
The comparison between HC and UASK (UA and SK patients pooled together) showed lower 360
FA values in the patients in all the tracts mentioned in the previous results, including the 361
middle cerebellar peduncle (Tmax = 5.14; k=1457; η2=0.10); the superior cerebellar peduncle 362
(Tmax = 5.20; k=278; η2=0.14); the fornix (Tmax = 7.20; k=609; η2=0.14) and the cingulum 363
(Tmax = 5.47; k=840; η2=0.14). Fig. 2 shows the white matter tracts of the FCC and PC that 364
are significantly disrupted in the UASK group compared to the HC group.
365
[FIGURE 2]
366 367
(ii) Identification of UA at risk to develop KS for each fibre tract 368
For all participants, mean FA values were extracted from significant fibre-clusters (see above) 369
for the middle and superior cerebellar peduncles, representing tracts from the FCC and the 370
cingulum and fornix, representing tracts from the PC. A k-means classification was then 371
performed on those tracts. Fig. 3 represents the distribution of the mean FA values for each 372
selected tract identified in the white matter atlas (Mori et al, 2010) to be part of the FCC and 373
PC.
374
With regard to white matter tracts within the PC, the k-means classification conducted on the 375
FA values in the cingulum and fornix showed an expected classification of all HC in one 376
group and all KS patients in another. Some UA patients were classified in the same group as 377
HC and can be therefore considered as UA_LOW, while others were classified in the same 378
group as KS and were thus qualified as UA_HIGH. The same patients were classified as 379
UA_HIGH for both the cingulum and the fornix, except for two of them (one for cingulum 380
and one for fornix). However, these 2 patients were those with the lowest FA values in the 381
2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59
For Peer Review
subsequent UA_LOW class. Comparisons of mean FA values among the 4 subgroups for both 382
the cingulum and the fornix (ANOVA Kruskal-Wallis) have shown that the controls did not 383
differ from the UA_LOW group (p=1). KS differed significantly from HC (p<0.001) and 384
UA_LOW (p=0.01) but not UA_HIGH. UA_HIGH was significantly different from HC 385
(p<0.001) and UA_LOW (p=0.01).
386
Regarding the tracts of the FCC, the results of the classification seemed less reliable since 387
each of the two groups identified included a mix of KS patients, UA patients and HC.
388
Moreover, the number and the identity of the UA patients classified as UA_HIGH completely 389
differed between the two tracts of the FCC.
390
[FIGURE 3]
391 392
(iii) Comparison of episodic memory performance in the UA_HIGH and UA_LOW groups 393
Since a better classification of patients for UA_HIGH and UA_LOW were obtained for the 394
two white matter tracts of the PC as opposed to the FCC, we compared episodic memory 395
performances between the two subgroups of UA identified with the classification analysis 396
conducted on FA values in the fornix and cingulum individually. Non-parametric Mann- 397
Whitney U tests showed that UA_LOW did not differ from HC (p=0.37 for the classification 398
using the fornix and p=0.32 for the classification using the cingulum). UA_LOW differed 399
from UA_HIGH (p=0.038 for the classification using the fornix and p=0.025 for the 400
classification using the cingulum). UA_HIGH differed significantly from HC (p=0.002 for 401
both fornix and cingulum) and KS (p=0.001 for classification with the fornix and p=0.002 for 402
classification with the cingulum). The latter was significantly different to all the other 3 403
subgroups for both classifications (p<0.001). These results are illustrated in Fig. 4.
404
[FIGURE 4]
405 2
3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59
For Peer Review
We conducted the same analysis on the performances on matrix reasoning, and forward and 406
backward block spans but found insignificant differences between the UA subgroups for 407
neither the cingulum nor the fornix (data not shown). Hence, the ability to identify UA 408
patients at risk of developing KS seems to be specific to performances in episodic memory 409
tasks.
410
Note that we have also tested classification with the corpus callosum but it yielded poor 411
results (data not shown since the corpus callosum is neither part of the FCC, nor the PC).
412 413
Discussion
414
The first aim of this study was to describe the profile of microstructural white matter 415
integrity in UA and KS in the whole brain, at a voxel-level, since most of the previous studies 416
had been carried out using a region of interest approach (with anatomical regions defined a 417
priori). In accordance with neuropathological (Harper et al. 2003; Kril et al. 1997) and 418
neuroimaging studies (Chanraud et al. 2010; Pfefferbaum and Sullivan, 2005; Pfefferbaum et 419
al. 2006, 2009), the present TBSS analysis of FA values in UA revealed widespread 420
compromised WM microstructure including notably fibers of the FCC and the PC. Using a 421
stringent statistical threshold (TFCE and FWE, p<0.05), our voxel-based analyses have 422
successfully replicated previous DTI results in fibres defined a priori (Pfefferbaum et al.
423
2009) such as the superior longitudinal fasciculus, external capsule, fornix and cingulum.
424
Moreover, our analyses revealed compromised WM integrity in other fibre tracts such as the 425
internal capsule, cerebral peduncles, corona radiata and thalamic radiations, which has also 426
been reported in a TBSS study comparing alcoholics who have been abstinent for at least 5 427
years with healthy controls (Fortier et al. 2014). Contrary to previous neuropathological 428
(Harper et al. 2003) and neuroimaging studies (Chanraud et al. 2009; Pfefferbaum et al.
429
2009), we also found microstructural abnormalities in the middle and superior cerebellar 430
2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59
For Peer Review
peduncles. The use of a voxel based approach may have enabled the observation of the latter 431
finding since a ROI approach, which would average the FA values within a region defined a 432
priori, would not reveal any localized impairments within those fibres. Thus, WM 433
abnormalities in the middle and superior cerebellar peduncles revealed by TBSS suggest that 434
disruption may be more localized than spread-throughout in those fibres.
435
Our study is the first to evaluate white matter microstructural integrity in KS patients.
436
When KS patients were compared to HC, the profile of WM abnormalities was similar to that 437
observed between UA and HC, in agreement with the patterns of GM and WM shrinkage 438
found in a recent VBM study (Pitel et al. 2012). Compromised WM microstructure in the 439
middle and superior cerebellar peduncles is in-line with previous investigations that have 440
shown significant shrinkage of the cerebellar white matter of KS patients (Harper et al. 2003;
441
Kril et al. 1997) and in proteomics studies where changes in the levels of thiamine-dependent 442
enzymes have been observed (Alexander-Kaufman et al. 2006).
443
Previous neuropsychological and neuroimaging (structural and functional) studies 444
have hypothesized that anterograde amnesia in KS patients is essentially due to a 445
disconnection within the PC (Warrington and Weiskrantz, 1982, Nahum et al. 2014). More 446
specifically, findings of abnormalities in diencephalic structures, including the thalamus and 447
mammillary bodies, and cortical structures from the frontal and medial temporal lobes have 448
pushed towards the hypothesis of a disruption between nodes belonging to this neural network 449
(Aupée et al, 2001; Renou et al. 2008; Kim et al. 2009). Our study provides consolidating 450
evidence of white matter disruptions in the PC that is linked to episodic memory deficits and 451
therefore amnesia (Kessels and Kopelman, 2012 for review). Based on the substantial role of 452
the cerebellum in cognitive processes and its connections with cortical areas, damaged FCC 453
has been hypothesized to be involved in working memory and executive dysfunction in KS 454
(Wijnia and Goossenssen, 2010). Our data confirm compromised WM integrity in the 455
2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59
For Peer Review
cerebellum and especially disruption of the middle and superior cerebellar peduncles. In 456
summary, cognitive deficits observed in KS patients, including amnesia, stems from a 457
disconnection of neural networks, which can in turn be due to abnormalities in the nodes of 458
the network(s), disruption of white matter tracts linking those nodes, or abnormal synaptic 459
activity between the nodes. While our study confirms the disruption of white matter tracts in 460
specific brain networks, and another study showed an absence of atrophy within connected 461
regions (Nahum et al, 2014), it is still difficult to evaluate the cascade of events 462
(neurotransmission dysfunction – local or global network disruption – cellular 463
damage/atrophy) that effectively governs the pathophysiological mechanism of KS when 464
using a cross-sectional paradigm,. Longitudinal studies are required to concretely establish 465
this mechanism.
466
Our direct voxel-based comparison between UA and KS patients showed significant 467
differences mainly in the corpus callosum, which follows previous volumetric studies that 468
has reported further volume loss in the corpus callosum in alcoholic patients with Wernicke’s 469
Encephalopathy than those alcoholic patients without (Lee et al., 2005) and the other one 470
between UA and KS (Pitel et al, 2012). The abnormalities in the microstructural integrity of 471
the corpus callosum have also been hypothesized to be due to thiamine deficiency as observed 472
in a study with rats (He et al, 2007).
473
Contrary to our initial hypotheses, a clear graded effect of deficits in the PC was not 474
observed. This can be attributed to our sample size and to the heterogeneity of the UA group.
475
As neurological complications from UA to KS lie along a continuum (Ryback, 1971), it is 476
difficult to observe a distinct pattern of microstructural degradation between these 2 groups, 477
justifying the need to divide the UA group into sub-groups to better observe their underlying 478
pathological mechanisms. By first combining the UA and KS groups, the statistical power to 479
detect all regions affected in alcohol dependent patients was increased (Monnig et al. 2013), 480
2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59
For Peer Review
ensuring that the subsequent classification step did not disregard any white matter tracts that 481
are potential structural biomarkers for identifying alcoholics at risk to develop neurological 482
complications such as KS.
483
The classification step revealed that the use of DTI may be particularly relevant as a 484
structural biomarker towards the early identification of UA patients at risk of developing KS.
485
The early identification is important for clinicians to apply the correct and optimal treatment 486
with the aim of preventing severe, debilitating and irreversible neurological complications.
487
Our analysis provides consolidating evidence of the PC, as opposed to the FCC, being an 488
appropriate neuroanatomical substrate for identifying UA patients that can potentially develop 489
KS (Pitel et al. 2012). The statistical comparisons of the mean FA values between the 490
subgroups of subjects have confirmed the robustness of the classification step. The fact that 491
there is a clear separation between HC and KS as well as consistent classification of the UA 492
subgroups for the cingulum and fornix complements volumetric findings that have shown 493
graded effects in the mammillary bodies, the hippocampus and the thalamus (Sullivan and 494
Pfefferbaum, 2009). In a previous study (Pitel et al. 2012), the volume of the thalamus was 495
found to be comparable between some of the UA patients and the KS ones, reinforcing 496
neuropsychological data that have shown the same trend in episodic memory deficits (Pitel et 497
al. 2008). While the volume of the thalamus was not explored in the present study, we 498
confirm that UA patients classified at risk to develop KS based on WM microstructural 499
abnormalities in the PC had the lowest episodic memory scores. Interestingly, it was also 500
observed that 4 patients out of 10 classified as UA_LOW filled none of Caine’s criteria for 501
Wernicke’s encephalopathy (Caine et al. 1997; Pitel et al. 2011), 4 patients filled one criterion 502
and 2 filled more than 2 criteria. For the UA_HIGH subgroup, 2 patients filled no criterion, 4 503
filled 1 criterion and 4 filled more than 2 criteria. The number of signs of Wernicke’s 504
encephalopathy did not differ between the UA_LOW and UA_HIGH (Chi-squared test, data 505
2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59
For Peer Review
not shown). Our refined neuroimaging analyses confirm that a neuropsychological evaluation, 506
especially targeting episodic memory, is highly recommended in clinical settings, where 507
neuroimaging tools are not available, to identify patients at risk of developing KS. Taken 508
together, the analysis of microstructural integrity within the fornix and cingulum, in 509
combination with scores of episodic memory, thalamic volume and signs of Wernicke’s 510
encephalopathy, could give a reliable depiction of whether a UA patient is at risk of 511
developing KS. Our study is thus a positive iteration to the heuristic value of the continuity 512
hypothesis (Butters and Brandt, 1985; Parsons, 1998; Pitel et al. 2008; Ryback, 1971). While 513
this inherent heterogeneity in the UA group enabled us to detect alcoholics at risk to develop 514
neurological complications, the currently used average neuropsychological, structural and 515
functional description of a UA group does not reflect the heterogeneity of individual profiles 516
in clinical settings.
517
518
Conclusions and further works
519
TBSS has allowed us to describe the profile of white matter integrity at a voxel-level in UA 520
and KS patients. While the chronological position of white matter disruption is still unknown 521
in the cascade of structural and functional events that govern the pathophysiological 522
mechanism underlying the neurotoxicity of alcohol, we have shown the potential of DTI data 523
to identify uncomplicated alcoholics at risk of developing KS. The method paves the way for 524
more in depth analysis of this subgroup in order to better understand the mechanism 525
underlying these two clinical forms. Multi-modal neuroimaging, combined with biological 526
and neuropsychological analyses will enable researchers to explore the characteristics of these 527
clinical forms in terms of detailed microstructure, regional volume, function, enzyme 528
metabolism and cognitive deficits. The specificity of the subgroup of patients at risk to 529
2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59
For Peer Review
develop KS is also likely to be confirmed via longitudinal studies in which the progress of the 530
pathology can be monitored and the efficiency of the treatment can be assessed and optimised.
531
532 2
3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59
For Peer Review
Figure legends:
533 534
Figure 1 Voxel by voxel comparisons of FA values between HC and UA (top row), HC and 535
KS (middle row) and UA and KS (bottom row) using non-parametric permutation tests (5000 536
permutations, FWE p<0.05, TFCE for cluster-wise correction). p-value maps are shown as (1- 537
p) images, displayed on a T1-weighted MRI in MNI space.
538 539
Figure 2 Voxel by voxel comparisons of FA values between HC and UASK using non- 540
parametric permutation tests (5000 permutations, FWE p<0.05, and TFCE for cluster-wise 541
correction). p-value maps are shown as (1-p) images, displayed on a T1-weighted MRI in 542
MNI space.
543 544
Figure 3 Distribution of the mean FA values for the selected white matter tracts within the 545
fronto-cerebellar circuit (mcp = middle cerebellar peduncle, scp = superior cerebellar 546
peduncle) and the Papez circuit (fornix and cing = cingulum). Horizontal black lines represent 547
the separation between the 2 identified clusters.
548 549
Figure 4 Episodic memory performance (sum of the 3 free-recalls of FCSRT) in the HC,KS 550
and subgroups UA_LOW and UA_HIGH.
551
UA_LOW and UA_HIGH have been identified using k-means classification conducted on 552
means FA values of significant fibre-clusters in the fornix and the cingulum bundle.
553
* : sig different from HC; § : sig different from UA_LOW; ¥ : sig different from UA_HIGH 554
2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59
For Peer Review
References
555
Aggleton JP, Brown MW (1999): Episodic memory, amnesia, and the hippocampal-anterior 556
thalamic axis. Behav Brain Sci. 22: 425–444.
557
Aggleton JP (2012): Multiple anatomical systems embedded within the primate medial 558
temporal lobe: Implications for hippocampal function . Neurosci Biobehav Rev. 36: 1579–
559
1596.
560
Alexander-Kaufman K, James G, Sheedy D, Harper C, Matsumoto I (2006): Differential 561
protein expression in the prefrontal white matter of human alcoholics: a proteomics study.
562
Mol psychiatry. 11: 56–65.
563
Ambrose ML, Bowden SC, Whelan G (2001): Working Memory Impairments in Alcohol- 564
Dependent Participants Without Clinical Amnesia. Alcohol Clin Exp Res. 25: 185–191.
565
American Psychiatric Association. Diagnostic and Statistical Manual of Mental Disorders. 4th 566
ed., Text Rev. Author, Washington, DC 2000.
567
Aupée AM, Desgranges B, Eustache F, Lalevée C, de la Sayette V, Viader F, Baron JC 568
(2001) Voxel-based mapping of brain hypometabolism in permanent amnesia with PET.
569
Neuroimage. 13(6): 1164-1173 570
Beck A, Ward C, Mendelson M, Mock J, Erbaugh J (1961): An inventory for measuring 571
depression. Arch Gen Psychiatry. 4: 561–571.
572
Butters N, Brandt J (1985): The continuity hypothesis: the relationship of long-term 573
alcoholism to the Wernicke-Korsakoff syndrome. Recent Dev Alcohol. 3: 207–226.
574 2
3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59
For Peer Review
Caine D, Halliday GM, Kril JJ, Harper CG (1997): Operational criteria for the classification 575
of chronic alcoholics: identification of Wernicke’s encephalopathy. J Neurol Neurosurg 576
Psychiatry 62: 51–60.
577
Chanraud S, Martelli C, Delain F, Kostogianni N, Douaud G, Aubin HJ, et al. (2007) : Brain 578
morphometry and cognitive performance in detoxified alcohol-dependents with preserved 579
psychosocial functioning. Neuropsychopharmacology. 32: 429–438.
580
Chanraud S, Pitel AL, Rohlfing T, Pfefferbaum A, Sullivan EV (2010): Dual tasking and 581
working memory in alcoholism: relation to frontocerebellar circuitry.
582
Neuropsychopharmacology. 35: 1868–1878.
583
Chanraud S, Reynaud M, Wessa M, Penttilä J, Kostogianni N, Cachia A, et al. (2009):
584
Diffusion tensor tractography in mesencephalic bundles: relation to mental flexibility in 585
detoxified alcohol-dependent subjects. Neuropsychopharmacology. 34: 1223–1232.
586
Chanraud S, Zahr N, Sullivan EV, Pfefferbaum A (2010): MR diffusion tensor imaging: a 587
window into white matter integrity of the working brain. Neuropsychol Rev. 20: 209–225.
588
Christie JE, Kean DM, Douglas RHB, Engleman HM, St. Clair D, Blackburn IM (1988):
589
Magnetic resonance imaging in pre-senile dementia of the Alzheimer-type, multi-infarct 590
dementia and Korsakoff’s syndrome. Psychol Med. 18: 319–329.
591
Colchester A, Kingsley D, Lasserson D, Kendall B, Bello F, Rush C, et al. (2001): Structural 592
MRI volumetric analysis in patients with organic amnesia, 1: methods and comparative 593
findings across diagnostic groups. J Neurol Neurosurg Psychiatry. 71: 13–22.
594
Fama R, Pitel AL, Sullivan EV (2012): Anterograde episodic memory in Korsakoff 595
syndrome. Neuropsychol Rev. 22: 93–104.
596 2
3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59
For Peer Review
Folstein MF, Folstein SE, McHugh PR (1975): Mini-mental state. A practical method for 597
grading the cognitive state of patients for the clinician. J Psychiatr Res. 12:189–198.
598
Gache P, Michaud P, Landry U, Accietto C, Arfaoui S, Wenger O, et al. (2005) : The Alcohol 599
Use Disorders Identification Test (AUDIT) as a Screening Tool for Excessive Drinking in 600
Primary Care: Reliability and Validity of a French Version. Alcohol Clin Exp Res. 29: 2001–
601
2007.
602
Grober E, Buschke H (1987): Genuine memory deficits in dementia. Dev Neuropsychol. 3:
603
13–36.
604
Harper CG, Kril JJ (1990): Neuropathology of alcoholism. Alcohol and alcoholism; 25: 207–
605
216.
606
Harper C, Dixon G, Sheedy D, Garrick T (2003): Neuropathological alterations in alcoholic 607
brains. Studies arising from the New South Wales Tissue Resource Centre. Prog 608
Neuropsychopharmacol Biol Psychiatry. 27: 951–961.
609
Harper C, Matsumoto I (2005): Ethanol and brain damage. Current Opinion in Pharmacology.
610
5:73-78.
611
Harper C (2007): The neurotoxicity of alcohol. Human and experimental toxicology. 26: 251- 612
257.
613
Harper C (2009): The neuropathology of alcohol-related brain damage. Alcohol and 614
alcoholism. 44: 136–140.
615 2
3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59
For Peer Review
Harris GJ, Jaffin SK, Hodge SM, Kennedy D, Caviness VS, Marinkovic K, et al. (2008):
616
Frontal white matter and cingulum diffusion tensor imaging deficits in alcoholism. Alcohol 617
Clin Exp Res. 32: 1001–1013.
618
He X, Sullivan EV, Stankovic RK, Harper CG, Pfefferbaum A (2007): Interaction of 619
Thiamine deficiency and voluntary alcohol consumption disrupts rat corpus callosum 620
ultrastructure. Neuropsychopharmacology. 32: 2207-2216 621
Heatherton TF, Kozlowski LT, Frecker RC, Fagerström KO (1991): The Fagerström Test for 622
Nicotine Dependence: a revision of the Fagerström Tolerance Questionnaire. Br J Addict.
623
86:1119–1127.
624
Ihara H, Berrios GE, London M (2000): Group and case study of the dysexecutive syndrome 625
in alcoholism without amnesia. J Neurol Neurosurg Psychiatry. 68: 731–737.
626
Jacobson RR, Acker CF, Lishman WA (1990): Patterns of neuropsychological deficit in 627
alcoholic Korsakoff’s Syndrome. Psychol Med. 20: 321–334.
628
Jenkinson M, Bannister P, Brady M, Smith S (2002): Improved Optimization for the Robust 629
and Accurate Linear Registration and Motion Correction of Brain Images. NeuroImage. 17:
630
825–841.
631
Jones DK, Symms MR, Cercignani M, Howard RJ (2005): The effect of filter size on VBM 632
analyses of DT-MRI data. NeuroImage. 26: 546–554.
633
Kelly RM, Strick PL (2003): Cerebellar loops with motor cortex and prefrontal cortex of a 634
nonhuman primate. J Neurosci. 23: 8432–8444.
635 2
3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59
For Peer Review
Kessels RP, Kopelman MD (2012) : Context memory in Korsakoff’s syndrome.
636
Neuropsychol Rev. 22(2): 117-131 637
Kim E, Ku J, Namkoong K, Lee W, Lee KS, Park J-Y et al. (2009) : Mammillothalamic 638
functional connectivity and memory function in Wernicke’s encephalopathy. Brain. 132: 369- 639
376.
640
Kopelman Michael D (1989): Remote and autobiographical memory, temporal context 641
memory and frontal atrophy in Korsakoff and Alzheimer patients. Neuropsychologia. 27:
642
437–460.
643
Korsakoff S (1889) : Etude medico-psychologique sur une forme des maladies de la mémoire.
644
Revue Philosophie. 20: 501–530.
645
Krabbendam L, Visser PJ, Derix MMA, Verhey F, Hofman P, Verhoeven W, et al. (2000):
646
Normal Cognitive Performance in Patients With Chronic Alcoholism in Contrast to Patients 647
With Korsakoff’s Syndrome. J Neuropsychiatry Clin Neurosci. 12(1): 44–50.
648
Kril JJ, Halliday GM, Svoboda MD, Cartwright H (1997): The cerebral cortex is damaged in 649
chronic alcoholics. Neuroscience. 79: 983–998.
650
Lee S-T, Jung Y-M, Na DL, Park SH, Kim M (2005): Corpus callosum atrophy in Wernicke’s 651
encephalopathy. J Neuroimaging. 15: 367-372 652
Van der Linden M (2004) : L’évaluation des troubles de la mémoire : Présentation de quatre 653
tests de mémoire épisodique. Collectif 2004.
654
Matsumoto I (2009): Proteomics approach in the study of the pathophysiology of alcohol- 655
related brain damage. Alcohol Alcohol. 44: 171-176.
656 2
3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59